CN101102115B

CN101102115B - Encoding and decoding method for packet information, encoder and decoder

Info

Publication number: CN101102115B
Application number: CN2007101088630A
Authority: CN
Inventors: 片山泰尚; 中野大树
Original assignee: International Business Machines Corp
Current assignee: International Business Machines Corp
Priority date: 2006-07-03
Filing date: 2007-06-05
Publication date: 2012-07-04
Anticipated expiration: 2027-06-05
Also published as: JP4392004B2; CN101102115A; US20080022182A1; US8037395B2; JP2008016907A

Abstract

The present invention discloses a Partially-Overlapped Block (POB) code used in a new encoding method. In this encoding method, a POB code is implemented by combining a plurality of different block codes so that the block codes partially overlap one another. A decoding method corresponding to this encoding method is also disclosed. In addition, disclosed is a method for recovering a plurality of packets by using a loss correction capability of this code. The present invention makes it possible to recover the larger number of packets than the number of added redundant packets per frame, by reusing redundant information of neighboring frames effectively, without increasing the asymptotic complexity of its decoding algorithm.

Description

The method of Code And Decode grouping information, encoder

Technical field

The packet recovery that relate generally to of the present invention uses in digital communication.More particularly, the present invention relates to be used for coding method, coding/decoding method, the encoder of packet recovery.

Background technology

For wired or wireless network application, be used for the coding techniques of packet recovery and more and more important with the corresponding decoding technique of this coding techniques.Its reason is that the increase of data transmission bauds often brings the problem of packet loss, and can increase performance cost and cache management expense based on the legacy packets restoration methods of packet retransmissions present.Here, packet loss is owing to information in the mistake in header/sync pattern detection etc. and the network switch temporary transient concentrated congested the causing that is caused.

As a kind of technology that solves foregoing problems, a kind of known method is to use such as the block code of the Read-Solomon that interweaves (RS) erasure correction sign indicating number and pursues frame ground to dividing into groups to encode (non-patent literature 1).But, in this conventional art, do not utilize redundancy packets, and just that this redundancy packets is discarded when in given frame, not having lost packets.This poor efficiency is set to can relax when longer at code length.But the so long complexity of calculation problem in the time of to cause decoding that is provided with of code length meanwhile.

As another example of conventional art, exist a kind of basis by transmitter and normally used source node of receiver and the graph structure that the inspection node constitutes, carry out the method (non-patent literature 2) of decoding for the specific non-systemic sign indicating number of packet recovery through use.This method is effectively aspect code rate, and owing to the time of substantial linear being adopted in decoding so also being efficient.But this method does not show good correction efficient, only if code length is equal to or greater than length-specific (being equal to or greater than specified quantitative).And the efficient of implementing this method through hardware is not high, and this is at random because of graph structure.For this reason, the range of application of this method is limited in the Streaming Media and data delivery of Code And Decode through the software realization.

As another kind of conventional art, (convolution code m) is carried out the method (non-patent literature 3) of packet recovery for n, n-1 to have a kind of use.In the method, lost packets can be recovered than the technology based on frame of using block code more neatly.Yet this method is based on Binary Convolutional Code, thereby the correction of losing that needs to pursue bit.For this reason, this method does not allow to carry out the decoding that is the basis with by symbol.Correspondingly, in order to implement this method, then a large amount of decoders need be set concurrently through hardware.

The aforementioned conventional technology has the problem that when improving the efficient of packet recovery, will cause long code length, decoder circuit realization complicacy etc.In addition, the redundant information that too much increases may cause new congested.Therefore, the efficient of improving packet recovery is very important challenge.

[non-patent literature 1] Nonnenmacher, E.W.Biersack, D.Towsley, " Parity-based loss recovery for reliable multicast ", IEEE/ACM Trans.Networking, vol.6, pp.349.361, in August, 1998.

[non-patent literature 2] Luby, M.Mitzenmacher, A.Shokrollahi, D.Spielman, " Efficient erasure correcting codes ", IEEE Trans.Inform.Theory, vol.47, pp.569.584, February calendar year 2001.

[non-patent literature 3] M.Arai, A.Yamaguchi, K.Iwasaki; " Method toRecover Internet Packet Losses Using (n, n-1, m) Convolutional Codes "; IEEE Dependable System and Network, pp.382-389,2000.

Summary of the invention

The purpose of this invention is to provide a kind of redundant information that when improving packet recovery speed, can efficiently utilize appropriate amount new coding method and with the corresponding coding/decoding method of this new coding method.In addition, the present invention also aims to provide a kind of new encoder and decoder that is used to realize said method.

The present invention discloses a kind of piece of overlapping (POB) sign indicating number that is used in the new coding method, and it constitutes through a plurality of block codes that combination differs from one another, thereby makes block code partly overlapping each other.In addition, the present invention discloses a kind of and the corresponding coding/decoding method of this coding method.In addition, the present invention discloses a kind of through using the method that calibration capability recovers a plurality of groupings of losing of this yard.The present invention is through utilize the redundant information of consecutive frame effectively again, makes it possible to recover be added with a large amount of groupings that the quantity of redundant grouping is Duoed than every frame, and need not to increase the asymptotic complexity of its encryption algorithm.

A kind of encoder that is used to produce POB sign indicating number of the present invention also is provided.This encoder comprises: the parts that are used to detect frame; And be used for through changing the parts of block code for each subframe generation parity check bit from one to another ground, wherein a plurality of block codes partly overlap each other.In addition, a kind of decoder is provided, comprises: the parts that are used to detect frame; The parts of subframe that are used for the quantity of the block code that stored number equals to overlap each other; Be used for through using different test matrix to produce the parts of syndrome for each sub-frame; Be used for confirming to be used for the parts of the pattern of error recovery through the information of using relevant packet error position and number of packet errors; The parts that are used for mistake in computation; The parts that are used for error recovery and decoding; And be used for through using the information of proofreading and correct to recomputate the parts of syndrome.

Description of drawings

Fig. 1 is the figure that the code structure of POB sign indicating number of the present invention is shown.

Fig. 2 is the figure that the example code structure of POB sign indicating number of the present invention is shown.

Fig. 3 is the figure that is illustrated in packet recovery effect under the situation of using POB coding method of the present invention.

Fig. 4 is the exemplary plot of POB method of the present invention.

Fig. 5 is illustrated in the structure chart in the state behind the generation parity check bit in the structure shown in Figure 4.

Fig. 6 is the exemplary plot under the situation that code word shown in Figure 5 has been decoded.

Fig. 7 is the figure that representes coding flow process in the present invention.

Fig. 8 illustrates the topology example figure of the encoder of the flow process of supporting Fig. 7.

Fig. 9 is the figure that representes decoding process in the present invention.

Figure 10 illustrates the topology example figure of the decoder 200 that is used to realize flow process shown in Figure 9.

Figure 11 illustrates through using the figure of the circuit scale that last public multiplier branch can reduce.

Figure 12 is illustrated in the structure chart when using RS sign indicating number of the present invention.

Figure 13 illustrates the flow chart of coding method shown in Figure 12.

Figure 14 illustrates the structure chart of the encoder that is used to realize flow process shown in Figure 13.

Figure 15 is illustrated in the flow chart of decoding under the RS sign indicating number situation.

Figure 16 illustrates the structure chart of the decoder that is used to realize flow process shown in Figure 15.

Figure 17 A and Figure 17 B illustrate each frame assumption diagram that comprises the concatenated code of the foreign key that is produced by the POB coding method.

Figure 18 is illustrated in the flow chart of the coding method under the situation of Figure 17.

Figure 19 illustrates the structure chart of the encoder of the flow process that is used to realize Figure 18.

Figure 20 is illustrated in the structure chart of the decoder under the situation of Figure 17.

Figure 21 illustrates the structure chart of the decoder that is used for internal code shown in Figure 20.

In the accompanying drawing:

10,30, the frame detection part;

12,32, the subframe detection part;

14,34, the detection of packets parts;

16,38,84, linear feedback register;

18,20,46,86, selector;

36, be used to detect the quantity of packet error and the parts of position;

40,44, register;

48, be used for the parts that the coefficient of equation is sought in mistake in computation;

50, parts are confirmed in error correction;

52, mistake produces (calculating) parts;

58, error register;

72, multiplier;

80, external encoder;

82, internal encoder;

100,300,500, encoder;

200,400,600, decoder.

Embodiment

Describe the present invention below with reference to accompanying drawings in detail.POB sign indicating number of the present invention uses code structure shown in Figure 1.Suppose that the code word w in each frame can be divided into the s sub-frame (wh) (supposition n and r can both be divided exactly by s) that comprises redundant symbol if structure comprises L time n the symbol of interweaving in the frame.In other words, each of subframe of supposing n/s symbol comprises r/s redundant symbol.

So, to the s that differs from one another (n, n-r/s) the nonbinary block code is encoded so that satisfy following parity check condition:

Σ_{h = j - s + 1}^{j} H_{ih} w_{h} = 0

Wherein

H _i＝[H _i(j-s+1)，H _i(j-s+2)，…，H _ij]，

H _ih＝[α ^ih*n/s，α ^i(h*n/s+1)，…，α ^{i(h*n/s+n/s-1)}]

H is the generator matrix; W is a code word; α is the primitive element in the Galois Field; S is the quantity (quantity of the subframe that equals to cut apart) of block code; N is the piece size, and k is an information size, and notes k=(n-r)/s; R is the parity check size; And j is a subframe numbers.Each sign indicating number not necessarily uses in present frame, but at the assemble of symbol wj of the j subframe s-to j subframe front1Assemble of symbol w in the sub-frameJ-s-1, wJ-s-2..., wJ-1Scope in use.In addition, i is used for the subfix of identification code condition and satisfies jr/s=<i<(j+1) rs, h*Be illustrated in the relative position (h in the h subframe of frame*=0,1,2 ..., s-1).

Decoding is carried out according to following step.It should be noted that at this for specific j subframe, if do not have packet error in the s-1 sub-frame in front, if perhaps the mistake in the s-1 sub-frame is all calibrated, it is similar that sign indicating number coding/decoding method of then in the disclosure, advising and block code are lost correction.At first, computing syndrome is then through using relevant information of losing the position to seek the symbolic information of relevant lost packets.Through using the data W of Hi and receptionJ-s-1, wJ-s-2..., wiThe i syndrome S at the end of j subframe is added in calculating toi(jr/s=<i<(j+1) rs).

S_{i} = Σ_{h = j - s + 1}^{j} H_{ih} w_{h}

For other mark that is added with *, such as j ^*, its definition mode and h ^*Situation identical.

If at its head is that the total quantity of lost packets in the h sub-frame of j subframe is less than p=rh/s, the grouping that then always can recover to lose.For example, if do not have lost packets in the h-1 sub-frame after current subframe, then can recover as many as p grouping.In other words, at q (q=<p) individual grouped data b _E0, b _E1..., b _Eq-1The e in the position respectively ₀, e ₁..., e _Q-1Under the situation of losing, grouping can recover through finding the solution following linear equation, and the situation that this and RS sign indicating number are lost correction is similar.

[\begin{matrix} S_{(jrjs) *} \\ S_{(1 + jrls) *} \\ S_{(2 + jrls) *} \\ \cdot \\ \cdot \\ \cdot \\ S_{(q - 1 + jrls) *} \end{matrix}] = [\begin{matrix} α^{(jrls) * e_{0}} & α^{(jrls) * e_{1}} & \cdot \cdot \cdot & α^{(jrls) * e_{q - 1}} \\ α^{(1 + jrls) * e_{0}} & α^{(1 + jrsl) * e_{1}} & \cdot \cdot \cdot & α^{(1 + jrls) * e_{q - 1}} \\ α^{(2 + jrls) * e_{0}} & α^{(2 + jrls) * e_{1}} & \cdot \cdot \cdot & α^{(2 + jrls) * e_{q - 1}} \\ \cdot & \cdot & \cdot \\ \cdot & \cdot & \cdot \\ \cdot & \cdot & \cdot \\ α^{(q - 1 + jrls) * e_{0}} & α^{(q - 1 + jrls) * e_{1}} & \cdot \cdot \cdot & α^{(q - 1 + jrls) * e_{q - 1}} \end{matrix}] [\begin{matrix} {be}_{0} \\ {be}_{1} \\ \cdot \\ \cdot \\ \cdot \\ {be}_{q - 1} \end{matrix}]

In case recovered all groupings of losing in the current subframe, then need recomputate syndrome on each border of subframe.This syndrome recomputates step and allows the maximum quantity of the correctable grouping of each frame to be equal to or greater than r, and prevents that the cost that calculates from increasing.The new syndrome Si ' that recomputates can calculate through following formula.

[\begin{matrix} {S'}_{0} \\ {S'}_{1} \\ {S'}_{2} \\ \cdot \\ \cdot \\ \cdot \\ {S'}_{r - 1} \end{matrix}] = [\begin{matrix} S_{0} \\ S_{1} \\ S_{2} \\ \cdot \\ \cdot \\ \cdot \\ S_{r - 1} \end{matrix}] + [\begin{matrix} 1 & 1 & \cdot \cdot \cdot & 1 \\ a^{e_{0}} & a^{e_{1}} & \cdot \cdot \cdot & a^{e_{q - 1}} \\ a^{{2 e}_{0}} & a^{2 e_{1}} & \cdot \cdot \cdot & a^{e_{q - 1}} \\ \cdot & \cdot & \cdot \\ \cdot & \cdot & \cdot \\ \cdot & \cdot & \cdot \\ a^{(r - 1) e_{0}} & \cdot \cdot \cdot & a^{(r - 1) e_{q - 1}} \end{matrix}] [\begin{matrix} {be}_{0} \\ {be}_{1} \\ \cdot \\ \cdot \\ \cdot \\ {be}_{q - 1} \end{matrix}]

Wherein, α is the primitive element in the Galois Field; b _E0, b _E1..., b _Eq-1Be relevant q (q=<p) grouping information of individual grouping; And e ₀, e ₁..., e _Q-1It is the positional information in dividing into groups.

As the total quantity h of the grouping of in subframe, losing during greater than p, the grouping that then can not recover to lose.In this case, should proofread and correct in order to restart the packet recovery operation (2s-1) individual follow-up subframe.Although the asymptotic complexity (asymptoticcomplexity) that is used for decoding in this case is different with not the having of block code that be used to decode, need change decoding algorithm according to the pattern of quantity wrong in the subframe in the practice.Therefore, it is extremely important to design suitable decoder circuit, so that make that the scale of decoder is real.Specifically, actual decoding calculating receives the position effects that the current subframe that just is being corrected is arranged in frame.Effective hardware of decoder realizes requiring careful consideration this point.The minimizing of the quantity of multiplier is more even more important for the reduction of circuit scale than the minimizing of multiplication number of times.

Fig. 2 illustrates the example code structure of POB sign indicating number.Symbolic frame (n=150) is divided into subframe (s=3, and each subframe comprises n/s=50 symbol).Then, through use 3 (150,149) RS sign indicating numbers (be 1 in the parity check condition respectively, α, α ²Situation under generate) recover to divide into groups.More specifically, the RS sign indicating number is expressed as following formula:

H0＝[1，1，...，1，1，1...，1，1，1，...，1]，

H ₁＝[α ¹⁰⁰，α ¹⁰¹...，α ¹⁴⁹，α ⁰，α ¹，...，α ⁴⁹，α ⁵⁰，α ⁵¹，...，α ⁹⁹]，

H ₂＝[α ⁰，α ²，...，α ⁹⁸，α ¹⁰⁰，α ¹⁰²，...，α ¹⁹⁸，α ²⁰⁰，α ²⁰²，...，α ²⁹⁸]

The data w in the j subframe wherein _jRelational expression below satisfying:

Σ_{h = j - 2}^{j} H_{ih} w_{h} = 0

Wherein subfix i representes employed parity check condition.I is along with relative position j in the j subframe ^*And change.For example, subfix i can change into i=j ^*

Table 1 show j, j+l, j+2 subframe and the quantity of the lost packets that can recover between corresponding relation.Value of symbol b _eBe to be positioned at the j of frame ^*The symbol that position e in the subframe that position (

position

0,1 or 2) is located is lost, and if desired can be through using about other loses position e ' and e " information calculate.

[table 1]

Situation	j	j+1	j+2
				A	1	-	-
B	2	0	-
				C	2	1	0
D	3	0	0

Situation A

Case B

Situation C

Situation D

b_{e} = \frac{α^{e'} α^{2 e ″'} S_{0} + (α^{e'} + α^{e' ″}) S_{1} + S_{2}}{(α^{e} + α^{e'}) (α^{e} + α^{e ″ ″})}

In situation C, e and e ' are arranged in same subframe, and e " be arranged in next subframe.

Fig. 3 illustrates effect of the present invention through topology example of the present invention more shown in Figure 2 and other example of conventional art.In Fig. 3, the packet loss probability when the x axle is illustrated in input, the y axle is illustrated in the packet loss probability after the packet recovery mechanism of using each type.At first, the conventional method with method of the present invention and use (150,147) RS sign indicating number compares.Though two kinds of methods are used identical bit rates, be 3 under the situation of conventional method for the maximum quantity of the recoverable grouping of each frame, and its maximum quantity is 5 under situation of the present invention.Secondly, the method with method of the present invention and use (50,48) RS sign indicating number compares.Through using the redundancy packets of half quantity in the conventional method, method of the present invention shows substantially the same recovery capability in the less zone of packet loss rate.

In addition, compare with the method for describing in the non-patent literature 2, method of the present invention is not used LDPC sign indicating number at random, and is to use the sign indicating number that has based on the structure of block code.Therefore, the suitable hardware of method of the present invention is realized, even and under the relatively short situation of code length, also show effect preferably.In addition, method of the present invention is compared with the method for using Binary Convolutional Code, method of the present invention makes it possible to proofread and correct by symbol ground, thereby on hardware is realized, can realize efficient preferably than the method for recovering by bit to divide into groups.

Coding/decoding method, the encoder of supporting POB sign indicating number of the present invention will be described below.At first use instantiation that the POB coding method is described once more.Fig. 4 illustrates that a frame comprises 3 sub-frame and a block code is applied to the situation of this 3 sub-frame.Use 3 block codes altogether at this.A frame comprises a plurality of groupings, the width (bit) and the vertical length of Fig. 4 is represented to divide into groups.Under the situation that the information with aforementioned structure is encoded, arrow 1 expression in the left and right sides is by first subframe to the, three subframes in the same number of frames of first block code covering; First subframe of second subframe that left and right sides arrow 2 expression is covered by second block code in the next frame; And second subframe of the 3rd subframe that left and right sides arrow 3 expression is covered by the 3rd block code in the next frame.In brief, 3 different block codes cover different zones respectively.As a result, arrow 1 to 3 is partly overlapping in the 3rd subframe for example.

Fig. 5 is illustrated in the structure shown in Figure 4 through generating the structure chart of parity check bit after encoding.More particularly, in Fig. 5, will be added to the 3rd subframe back, will be added to first subframe back corresponding to the parity check bit of arrow 2, and will be added to second subframe back corresponding to the parity check bit of arrow 3 corresponding to the parity check bit of arrow 1.Needless to say more, being used to add the parity check bit position can limit arbitrarily.

Fig. 6 is the figure that is used to explain decodes codeword shown in Figure 5.Left and right sides arrow 1 illustrates from the zone of the parity check bit of first subframe to the, three subframes back.The zone of the parity check bit of first subframe back left and right sides arrow 2 illustrates from second subframe to next frame.The zone of the parity check bit of second subframe back left and right sides arrow 3 illustrates from the 3rd subframe to next frame.The syndrome of each of these zones can be calculated through the corresponding correction matrix of block code that uses the zone of representing with the respective digital that is used for Fig. 4.Thus, be that a sub-frame generates 3 syndromes, and each syndrome comprise the information of relevant this subframe.Then, use these 3 syndromes to carry out packet recovery through decodes codeword.

Under the situation of traditional block code, will be applied on the information in the same area such as those block codes by arrow 1 to 3 expression.On the other hand, under situation of the present invention, moved by block code 1 to 3 region covered, parity check bit overlaps each other simultaneously, and is as shown in Figure 6.This can improve with the frame is the error correction capability on basis.More particularly, exist and to comprise about the information of same sub but cover the block code of zones of different respectively.As a result, if in two sub-frame subsequently, do not have mistake, then can open up the maximum error correction capability of first subframe.The parity information that this means the inerrancy subframe can be used for the error correction of subframe subsequently effectively.Under the situation that mistake can be corrected, recomputating of syndrome causes the effectively further enhancing of application.In addition, as concrete coding/decoding method, the similar method of method of the block code of the current use of can using and decode.In this case, decoding needs according to the subframe numbers that will proofread and correct and comprises in correlation-corrected how much wrong situations are confirmed.

In the time can additionally receiving or detect packet error information, the POB coding method is unusual efficient coding method.Packet error information is to be illustrated in the frame to have mistake and specifically show the position of packet error and the information of quantity to be grouped into the basis.This information about position and quantity can be generated as sign set when the detection of packets parts can not correctly be discerned grouping, will be described below.Alternatively, also can receive by before decoder, handling the information that other system detected of dividing into groups about position and quantity.This position and quantity can receive or detect with packet synchronization ground.For this reason, packet error information calculations parts can the stores packets errors presents and are calculated the quantity of packet error in subframe.

< structure of encoder >

Fig. 7 illustrates coding flow process in the present invention.At first, receiving with the frame is the signal (7-A) that send on the basis.From the signal that is received, detect frame information (7-B).After detecting frame, detect subframe (7-C).Here, form subframe through the section that frame is divided into predetermined quantity.Detect the grouping (7-D) that constitutes subframe.Then, generate the parity check bit (7-E) of the quantity of the block code that and the block code that differ from one another relevant with subframe and quantity equals to be prepared concomitantly.At last, output parity check bit and the initial information (7-F) that generates like this.

Fig. 8 illustrates the example structure of the encoder 100 of supporting flow chart shown in Figure 7.Show that in head the head of frame is included under the situation in the information, frame detection part 10 detects head through pattern matching.Can extra acquisition and the situation of the synchronous control signal of the head of frame under, frame detection part 10 uses these control signals.Subframe detection part 12 and detection of packets parts 14 are to work with frame detection part 10 identical modes.Particularly, when frame structure is confirmed in advance, also can detect subframe and grouping through the synchronous counter of head of use and frame.Owing to a sub-frame a plurality of different block codes of encoding, therefore need to use the parity check with the number of employed block code to generate parts with overlapping mode.But, should be noted that the POB coding method does not also rely on the mode of selecting block code.As concrete circuit structure, need with the number of bit rate in the parity check bit that generates according to selected block code or symbol rate generate the circuit of parts as parity check.Fig. 8 shows these circuit that each parity check generates parts with overlaying state.Processing in these circuit of each parity check generation parts is carried out simultaneously.Generate parts as parity check, each quantity of register 16 that all has feedback unit is identical with the quantity of the block code of being supported.Feedback register 16 is arranged concurrently.Can begin or last begins from it from first of feedback register 16 to the input of feedback register 16.The structure of feedback register 16 is confirmed according to the type of the block code that will use.The result that subframe detects sends to the selector 18 that is connected with the end of feedback register 16.Then, confirm which subframe back what parity check is created on.Acting selector 20 is used for arbitrary format output information and parity check bit in the step in the end.

< structure of decoder >

The decoder that is used for receiving the decode in POB coding method information encoded of the present invention comprises the structure that for example has following parts.

1. frame detection part.

2. subframe detection part.

3. detection of packets parts.

4. syndrome computations parts are through using and the corresponding test matrix computing syndrome of each block code that in coding, uses.Block code has nothing in common with each other for the block of information of and a plurality of receptions that will be encoded relevant with subframe.

5. syndrome recomputates parts.Can be on correction time under the situation of mistake in some subframes in the subframe front of particular attention given, syndrome recomputates the corresponding correct syndrome of subframe that parts recomputate and seek and paid close attention to.

6. be used to detect or receive the parts of packet error information of the position of the quantity that shows packet error and packet error.

7. the wrong parts that generate.At first, mistake generation parts confirm according to the position of subframe and the quantity of packet error whether mistake can be corrected.Then, when mistake can be corrected, mistake generated the syndrome generation error that parts recomputate through use.

8. the decoding parts are used under the situation that mistake can be corrected, correctly the mistake that is generated being decoded.

9. output block.

Fig. 9 illustrates the decoding process among the present invention.Figure 10 illustrates the decoder 200 of supporting flow process shown in Figure 9.After receiving signal (9-A), at first detect frame (9-B).Then, detect subframe (9-C).Detect the grouping (9-D) that constitutes each subframe.If detect packet error, then with the quantity of packet error and location storage in packet error register 36, and the quantity of packet error in the subframe counted.These are input to and are used to the wrong parts 48 (hereafter is coefficient calculations parts 48) of seeking the Equation for Calculating coefficient and are used for the error correction that can definite mistake be corrected and confirm parts 50 (seeing also Figure 10).In addition, be the block of information computing syndrome (9-E) relevant through using with subframe with the corresponding test matrix of the block code of in coding, using.Because the information in the sub-frame is included in a plurality of block codes, the quantity of employed block code is identical under the situation of the quantity that therefore needs the syndrome computations parts and parity check bit generation when the coding.Computing syndrome concomitantly.As described in the structure of encoder, the POB coding method does not rely on the selection mode of the block code that will use.For this reason, a plurality of syndromes can be used for a block code.Therefore, as concrete circuit structure, need the quantity of circuit identical with the quantity of the syndrome of in a block code, using.

After computing syndrome, need recomputate some syndrome (9-F).The syndrome that recomputates is the syndrome in the error recovery in the information of those subframes that comprise relevant concern and the subframe that is used in the last subframe front of paying close attention to of time.It is under the situation that mistake can be corrected, to carry out that this syndrome recomputates.Formerly state (9-F) after all syndromes of information that mode recomputated each subframe that all comprises the relevant subframe front of paying close attention to, calculate mistake (9-G) through using these syndromes.Then, confirm that can mistake be corrected (9-I).If mistake can not be proofreaied and correct, sign is set is used for announcement information and comprises mistake (9-J).If mistake can be corrected, then calculate mistake, execution error is proofreaied and correct and decoding (9-K) then.At this moment, if mistake can be corrected, as stated, the error feedback that will in correction, use is returned the syndrome of when subsequently subframe is correct, carrying out and is recomputated.At last, decoded result is from output block output (9-L).Should be noted that, there is no need the output error sign.

Decoder 200 shown in Figure 10 comprises frame detection part 30, subframe detection part 32 and detection of packets parts 34, and is the same with the situation of encoder shown in Figure 8.Syndrome computations parts 38 are each grouping computing syndrome.Can use linear feedback shift register to form with the corresponding test matrix of the block code of in coding, using.Correspondingly, as the syndrome computations parts, linear feedback register 38 needs that are used for computing syndrome are prepared the quantity identical with the quantity of the block code that differs from one another.In addition, the information stores of input is in buffer 44.A plurality of syndromes are stored in the syndrome register 40.Syndrome recomputates (42) through the error correction result of the subframe of use front.Then, correct syndrome is input to coefficient calculations parts 48.Here, data can serial or parallel ground send to syndrome from syndrome register 40 and recomputate (42).

On the other hand; Detect under the situation of packet error at detection of packets parts 34; Perhaps receive under the situation about the information of packet loss at the circuit of decoder front, the position of stores packets mistake (36), and simultaneously the quantity of packet error is counted (36).Information about the packet error position is input to coefficient calculations parts 48.The syndrome computations coefficient that coefficient calculations parts 48 recomputate through the information combination of using input.The quantity of packet error is input to error correction to be confirmed in the parts 50, and error correction confirms that can parts 50 definite mistakes be corrected.The structure of confirming to depend on the POB sign indicating number that can relevant mistake be corrected is as described in the hereinafter.Under the situation that mistake can be corrected, mistake generates parts 52 mistake in computations.As a result, the information that is stored in the buffer 44 is corrected (54), then output.At this moment, if there is mistake, then about the information stores of mistake in error register 58.Then, as stated, feed back to (60) syndrome about the information of mistake through selector 46 and recomputate, so that the mistake in the correct subsequent subframe.Should be noted that decoder shown in Figure 10 can have error register 58 is foreclosed and is stored in the structure that the information in the syndrome register 40 is upgraded through result's feedback of using syndrome to recomputate (42).Thus, can the reduction circuit scale.

Here will further provide the description of the situation that can be resumed like above-mentioned table 1 grouping.Have 3 parity check bits that overlap each other for a sub-frame and existing under the situation of packet error positional information, each subframe can be recovered 3 groupings at the most.But, only when recovering 3 groupings (situation D) in the subframe when not having mistake in first subsequent subframe (j+1) and second subsequent subframe (j+2).Situation A to D representes the combination of wrong correctable situation.For example, if in current subframe, only there is a mistake, then can proofread and correct this mistake and irrelevant with the number of errors in other subframe.In addition, in these four kinds wrong recoverable situations, the coefficient that is used for generation error is according to the sub-frame number of the current subframe of frame front and difference.Since at the most 3 wrong have 3 syndromes, therefore need find the solution 3 equations simultaneously.Can obtain answer through the equation among each situation A to D of analysis and solution.In addition, coefficient is according to the sub-frame number of current subframe front in the frame and difference.Through using wrong quantity, confirm to be suitable for which kind of situation.Then, calculate the wrong coefficient of seeking.

The location of mistakes is sought coefficient and is related to the calculating of molecule and the calculating of denominator.Particularly, the calculating of the inverse of denominator needs large-scale circuit.For this reason, be used for the circuit that molecule and denominator reciprocal multiply each other through shared calculating molecule and the circuit that calculates denominator reciprocal and through shared, circuit scale can obviously descend.Figure 11 illustrates the example that circuit scale can descend thus.In addition, syndrome is only relevant with the calculating of molecule, so the input of syndrome can send to the shared part of molecular computing.

< the RS sign indicating number is as the situation of block code >

Often can be as the block code of POB sign indicating number as the reed-solomon code (RS sign indicating number) of nonbinary block code.Specifically, packet fragmentation is become to have separately the section of the symbol intervals of in the RS sign indicating number, using, application of RS sign indicating number then is so that the section of the overall width with the width that equals to divide into groups of interweaving.Polytype sign indicating number is prepared as the RS sign indicating number, and dissimilar RS sign indicating numbers is applied to zone separately, each is regional relevant with subframe and overlap each other.Thus, generate parity check bit.

Structure when Figure 12 illustrates the application of RS sign indicating number.A packet fragmentation is become to have separately the section of the symbol intervals of in the RS sign indicating number, using.To comprising that each grouping information with the zone shown in the section of such symbol intervals and in the arrow 1 to 3 of coverage diagram 4 any one carries out the calculating of using the RS sign indicating number, and generate parity check bit thus.Here, the quantity of cutting apart of subframe is arbitrarily, and the structure of RS sign indicating number also is arbitrarily.Correspondingly, can generate a plurality of parity check bits for a sub-frame.For example, shown in figure, when a sub-frame is divided into 3 sub-frame and (n, n-1) sign indicating number generates a parity check when the RS sign indicating number.Here, as the α of the base of relevant Galois Field in order to obtain 3 different RS sign indicating numbers.Specifically, with grouping information multiply by 1, α and α ², addition then.Because 3 RS sign indicating numbers that so obtain differ from one another, so the POB sign indicating number can be made up of these 3 RS sign indicating numbers.Specifically, will have 1 RS sign indicating number and be applied in the frame, will have α and be applied to the zone of first subframe of second subframe in the next frame from a frame, and will have α as the RS sign indicating number of root from the zone of first subframe to the, three subframes as root ²Be applied to the zone of second subframe of the 3rd subframe in the next frame from a frame as the RS sign indicating number of root.

Figure 13 is illustrated in the flow chart of the coding method in this situation.Figure 14 illustrates the structure of the encoder 300 of this method of support.In Figure 13, signal receives (13-A), frame and detects (13-B), subframe to detect (13-C), detection of packets (13-D) and output (13-F) identical with shown in Fig. 7 those.In Figure 14, represent with identical Reference numeral with assembly identical among Fig. 8.Prepare 3 RS sign indicating numbers that differ from one another.The RS sign indicating number uses the RS sign indicating number as block code, and adapts to Galois Field GF (2 through using ^m) primitive element α have 1 respectively, α, α ²The sign indicating number root.Through using the RS sign indicating number, relevant with subframe and by RS sign indicating number region covered in be all subframe generation parity check bits (13-E).Parity check bit can be arranged arbitrarily.Then, these are exported as coding result.Be used for 1, α, α ²Multiplier 70 append to feedback fraction respectively at the linear feedback shift register 16 of this use.Thus, shift register 16 can be through receiving and using the input information that always after multiply by fixed numbers, obtains to come computation of parity bits.But, should be noted that 3 shift registers are handled a plurality of subframes that its head is displaced to next subframe in proper order.For this reason, the subframe testing result is connected to and is positioned at the terminal selector 18 of shift register, and confirms which parity check output is in which subframe back.The selector 20 that uses in the step in the end is used for the parity check bit of output information and arbitrary format.

Below decoder will be described.Decoder receives the information of adding upward different with 3 the corresponding parity check bits of RS sign indicating number of being sent, and generates syndrome through using with the corresponding test matrix of each RS sign indicating number.Figure 15 illustrates the flow process of this processing.Figure 15 is identical with Fig. 9 basically.Here suppose that 3 RS sign indicating numbers have yard root 1, α, α respectively ², and the syndrome that obtains from these parity check bits is s0, s1 and s2.In this case, through using these parity check bits and calculating mistake about the information of packet error position.Because a packet fragmentation become a plurality of symbols, so the syndrome that obtains of each frame quantity cut apart of is-symbol at the most.

Figure 16 illustrates the structure of the decoder 400 of supporting flow chart shown in Figure 15.Structure shown in Figure 16 is basically with shown in Figure 10 identical.Under the situation of using the RS sign indicating number, the syndrome computations circuit can be configured to feedback register 38 and be connected to and be used for yard root 1, α, α ²Multiplier 70.Syndrome is stored in the register, and the error correction result of the subframe of time processing recomputates syndrome (42) through using in front.Then, the syndrome that recomputates is input to coefficient calculations parts 48.In addition, reception shows the quantity of packet error and the packet error information (36) of packet error position, and will be input to coefficient calculations parts 48 about the information of packet error position.As shown in Figure 8, the syndrome that can recomputate through use and come mistake in computation about the information of packet error position.In addition, select the pattern of wrong recoverable situation through the quantity of using packet error, as shown in table 1.The wrong equation of seeking is set up in position according to the pattern of situation A to D in the selected table 1 and the subframe that will proofread and correct.Then, generate the coefficient that is used for this equality.At last, through multiplier 72 mistake in computations shown in the afterbody of Figure 11.In case calculated mistake, through mistake being multiply by based on the coefficient of packet error position and subframe position and through resulting mistake is added to the information that is stored in the buffer 44 to come mistake is proofreaied and correct.This calculating is carried out to all symbols in dividing into groups, and output decoder result (56).If the error correction success is then returned error feedback to syndrome through error register 58 and is recomputated circuit (42), for use in proofreading and correct next subframe.

< encoder in the POB coding method of using concatenated code >

Although the POB coding method is unusual efficient coding method for packet recovery, use as foreign key if will comprise the concatenated code of POB sign indicating number, then this method can be more effective.In this case, can elect internal code as such as any sign indicating number of convolution code, and internal code mainly is configured to block code (Hamming code, BCH code, RS sign indicating number etc.) as a rule.Figure 17 is illustrated in the frame assumption diagram that uses the concatenated code that comprises foreign key in the POB coding method.A block code is applied in the grouping, and adds parity check bit the end of grouping to.For this reason, if will regard a grouping as to its code word of using internal code, the coding method that then is used for the POB sign indicating number is identical with foregoing POB coding method.Correspondingly, in this case flow chart and circuit structure further comprises the encoder that is used for as the block code of internal code.Through adding this internal code, can expect to improve calibration capability to the random error in dividing into groups.In addition, this method makes packet error to extract reliably, and this makes it possible to be implemented in the less circuit scale that uses the decoder of foreign key in the POB coding method.In addition, all essential processing can be carried out in the circuit structure inside of sealing, and need not the reception packet error information from the outside.

Figure 18 illustrates flow chart, and Figure 19 illustrates circuit structure 500.In Figure 18, signal receives (18-A), frame and detects (18-B), subframe to detect (18-C), detection of packets (18-D) and output (18-G) identical with shown in Fig. 7 those, and is widely different but parity check bit generates partly (18-E and 18-F).Generating foreign key, but generate the internal code that is used for a grouping concomitantly with the identical mode of the described mode of Fig. 7.Although shown in Figure 180 these are handled concurrently arrange, also can use the method for serial process, wherein at first carry out the processing that is used for internal code, carry out the processing that is used for foreign key then.But; Because the message part that divides into groups not only is used to generate the parity check bit of foreign key; And be used to generate the parity check bit of internal code, thereby the method that the method for carrying out these processing concurrently shown in figure 18 is carried out these processing than serial ground is more superior aspect the processing speed.In addition, in the part that the parity check bit of internal code and foreign key overlaps each other, be to use the processing of internal code to distribute high priority.In other words, at first generation is used for the parity check bit as the POB sign indicating number of foreign key, generates internal code to this parity check bit then.At last, output encoder result (18-G).

Basic structure shown in Figure 19 is with shown in Figure 8 identical.In Figure 19, this structure generates part in parity check and has the identical distinguishing characteristics of situation with the flow chart of Figure 18.As described in the description of flow chart, be used for parity check as the POB sign indicating number of foreign key and generate parts 80 and generate parts 82 with the parity check that is used for internal code and carry out their processing concurrently.As the encoder that is used for internal code, employed is linear feedback register 84, identical with the linear feedback register 16 of the encoder that is used for foreign key.According to situation, when using block code, use linear feedback register 84 to foreign key.

Figure 20 illustrates the example structure of decoder 600.In encoder, the encoder that is used for internal code can perhaps be arranged to serial with the encoder that is used for foreign key concurrently.But under the situation of decoder, the decoder 90 that is used for internal code is arranged with the decoder 92 serials ground that is used for foreign key.The decoder 90 that is used for internal code to be being grouped into based process information, and can adopt the general coding/decoding method to block code, and this is because grouping is through using arbitrarily block code to encode.But, existing under packet error and the wrong situation that can not be corrected, the decoder 90 that is used for internal code detects the position of packet error.In addition, the decoder 90 that is used for internal code also detects the number of packet errors how many packet errors expression exists, and exports this position and quantity.The decoder 92 that is used for foreign key in the POB coding method receives and divides into groups and packet error information, and through utilizing these to carry out decoding.

< decoder that is used for internal code >

Figure 21 illustrates the decoder 700 that is used for internal code.Block code is being used as under the situation of internal code, also can coming computing syndrome through combining linear feedback register 38.Internal code is through being usually used in proofreading and correct the purpose of random error in the grouping.For this reason, be used to proofread and correct the block code of a plurality of bit-errors as a rule.Thereby,, prepare a plurality of parity check bits owing to can proofread and correct a plurality of mistakes.The syndrome of the corresponding quantity of quantity of calculating and parity check bit is stored in the syndrome register 40 then.Calculate the coefficient of error location polynomial (48), the location of mistakes then through using calculated syndrome like this.If mistake can be corrected, the decoder that then is used for internal code is through using the mistake of the error correction grouping of so finding, input should be divided into groups then.If mistake can not be corrected, the decoder that then is used for internal code is not carried out any correction and is exported this grouping.In addition, the decoder output that is used for internal code shows sign that mistake can not be corrected and about the information in the packet positions of subframe.In other words, except as the packets of information that receives, also export two types information together, i.e. packet error sign and packet error position are as packet error information.In addition, Figure 21 illustrates buffer register 44.Under the situation of the decoder that is used for internal code, the size of buffer can be only the same big with the size of dividing into groups.The decoder that is used for internal code can adopt the general solution code calculation to employed block code.The distinguishing characteristics of concatenated code and POB sign indicating number is to generate the packet error sign, and extracts and the errors present of output in subframe.

< decoder that is used for foreign key (POB sign indicating number) >

As stated, the grouping of correction and packet error information are input to the decoder that is used for foreign key from the decoder that is used for internal code.Therefore, be input in the foreign key decoder about the information of the quantity of the packet error that in block code institute region covered, exists and about the information of the position of packet error.Correspondingly, the decoder that is used for foreign key can use the structure identical with structure shown in Figure 10 to realize.

So far, through using the embodiment shown in Fig. 1 to 21 to describe the present invention.But the present invention is not limited only to these embodiment.Obviously, those skilled in the art can carry out various other changes and modification to embodiment under the situation that does not break away from the spirit and scope of the present invention.

Claims

1. the method for a coding groups information comprises step:

Preparation comprises the frame of a plurality of groupings;

Form a plurality of subframes of forming said frame through cutting apart said frame; And

Form each and at least two corresponding block codes of adjacent sub-frame, overlap each other at least two different block codes of each subframe.

2. the method for a coding groups information comprises step:

Obtain a plurality of successive frames, each frame comprises a plurality of groupings; And

Form each with at least two frames in form the corresponding block code of at least two adjacent areas of said frame, overlap each other at least two the different block codes in each zone in the frame.

3. the method for a coding groups information comprises step:

Detection comprises the frame of a plurality of groupings;

Detect the subframe of the said frame of composition in the detected frame;

Detect the grouping in each subframe; And

Form at least two different block codes corresponding with each subframe and that overlap each other.

4. according to any described method in the claim 1 and 3, wherein block code forms step and also comprises step: with parity check bit add to the corresponding subframe of each block code in one on.

5. method according to claim 2, wherein block code forms step and also comprises step: with parity check bit add to the corresponding frame of each block code in the zone in one on.

6. according to any described method in the claim 1 and 3, wherein:

Block code is a reed-solomon code; And

To different block codes (n, k) encode to satisfy following formula:

Σ_{h = j - s + 1}^{j} H_{ih} w_{h} = 0

Wherein

H _i＝[H _i(j-s+1)，H _i(j-s+2)，…，H _ij]，

H _ih＝[α ^ih*n/s，α ^i(h*n/s+1)，…，α ^{i(h*n/s+n/s-1)}]

H is the generator matrix; W is a code word; α is the primitive element in the Galois Field; S is the quantity of block code, wherein the quantity of the s subframe that equals to cut apart; N is the piece size; K is an information size, and satisfies k=(n-r)/s; R is the parity check size; J is a subframe numbers; I is used for the subfix of identification code condition and satisfies the r/s of jr/s=＜i＜(j+1); And h* is the relative position of h subframe in frame, h*=0 wherein, and 1,2 ..., s-1.

7. the method for a coding groups information comprises step:

Detection comprises the frame of a plurality of groupings;

Detect the subframe of the said frame of composition in the detected frame;

Detect the grouping in each subframe;

Through using with corresponding at least two the different block codes generation parity checks of each subframe to come foreign key is encoded, wherein overlap each other at least two different block codes of each subframe; And

Each corresponding parity check of dividing into groups through in generation and the block code comes internal code is encoded.

8. the method for a decoded packet information comprises step:

Detection comprises the frame of a plurality of groupings;

Detect the subframe of the said frame of composition in the detected frame;

Detect the grouping in each subframe;

For with corresponding each the block code computing syndrome of a plurality of subframes, at least two different block codes of each subframe correspondence wherein, and overlap each other at least two different block codes of each subframe; And

Response whether can proofread and correct with the corresponding a plurality of subframes of previous calculated syndrome in, packet error on the time in the subframe of specific sub-frame front, recomputate the syndrome in the specific sub-frame.

9. method according to claim 8, wherein the syndrome computations step also comprises step: calculate the syndrome with the number of different block codes.

10. method according to claim 8, wherein:

Block code is a reed-solomon code; And

In the syndrome computations step, use computes syndrome Si

S_{i} = Σ_{h = j - s + 1}^{j} H_{ih} w_{h}

Wherein

H _i＝[H _i(j-s+1)，H _i(j-s+2)，…，H _ij]，

H _ih＝[α ^ih*n/s，α ^i(h*n/s+1)，…，α ^{i(h*n/s+n/s-1)}]

H is the generator matrix; W is a code word; α is the primitive element in the Galois Field; S is the quantity of block code; N is the piece size; I is a syndrome number, and j is subframe numbers, and satisfies the r/s of jr/s=＜i＜(j+1), and wherein r is the parity check size, and h* is the relative position of h subframe in frame, h*=0 wherein, and 1,2 ..., s-1.

11. method according to claim 10 is wherein used computes syndrome S ' in the syndrome computations step:

[\begin{matrix} {S^{'}}_{0} \\ {S^{'}}_{1} \\ {S^{'}}_{2} \\ . \\ . \\ . \\ {S^{'}}_{r - 1} \end{matrix}] = [\begin{matrix} S_{0} \\ S_{1} \\ S_{2} \\ . \\ . \\ . \\ S_{r - 1} \end{matrix}] + [\begin{matrix} 1 & 1 & . . . & 1 \\ α^{e_{0}} & α^{e_{1}} & . . . & α^{e_{q - 1}} \\ α^{{2 e}_{0}} & α^{{2 e}_{1}} & . . . & α^{{2 e}_{q - 1}} \\ . & . & . \\ . & . & . \\ . & . & . \\ α^{(r - 1) e_{0}} & α^{(r - 1) e_{1}} & . . . & α^{(r - 1) e_{q - 1}} \end{matrix}] [\begin{matrix} {be}_{0} \\ {be}_{1} \\ . \\ . \\ . \\ {be}_{q - 1} \end{matrix}]

Wherein, α is the primitive element in the Galois Field; b _E0, b _E1..., b _Eq-1Be q grouping information, and q=＜p, wherein p=rh/s; And e ₀, e ₁..., e _Q-1Position during expression is divided into groups.

12. an encoder comprises:

Be used to detect the parts of the frame that comprises a plurality of groupings;

Be used for detecting the parts of subframe of the component frame of frame;

Be used for detecting the parts of the grouping of each subframe; And

Be used to generate the parts with the corresponding parity check of subframe, each in the parity check wherein overlaps each other for two different block codes of each subframe corresponding to at least two different block codes of each subframe each at least.

13. encoder according to claim 12, wherein parity check generation parts comprise and the corresponding linear feedback register of each block code.

14. encoder according to claim 13 comprises that also output is through generating the parts of the coding result that parity check obtained;

Wherein, output block comprises selector, is used for output and the corresponding parity check of each sub-frame, switches to another from a parity check simultaneously.

15. a decoder comprises:

Be used for detecting the parts of subframe of the component frame of frame;

Be used for detecting the parts of the grouping of each subframe;

Be used to the parts that generate syndrome with corresponding each block code of a plurality of subframes, wherein overlap each other at least two different block codes of each subframe; And

Be used for responding whether can proofread and correct with the corresponding a plurality of subframes of the syndrome of previous generation, packet error on the time in the subframe of specific sub-frame front, regenerate the parts of the syndrome in the specific sub-frame.

16. decoder according to claim 15 also comprises:

Be used for through using the syndrome that is generated to detect the parts of the wrong position of grouping;

Be used for through using the syndrome that is generated to generate the parts of the mistake of dividing into groups; And

Be used for mistake, output decoder result's parts through the grouping of using the wrong position of the grouping detected and being generated.

17. decoder according to claim 15, wherein syndrome generation parts comprise:

Linear feedback register is used for generating and each corresponding parity check at least two different block codes of each subframe; And

Register is used to store the syndrome that is generated.

18. decoder according to claim 17; Wherein syndrome regenerate parts through be used in a plurality of subframes corresponding to the syndrome that is generated, the time tightens the mistake that the subframe that is connected on the specific sub-frame front generates, and regenerates the syndrome in the specific sub-frame.